Abstract
While Artificial Intelligence (AI) has demonstrated significant potential in the divergent phases of design, its role in the convergent phase (solution selection) is less understood. This study investigates how different AI-assisted selection modes impact a designer's creative efficacy. We conducted an experiment (N=40) with graduate design students, who used four distinct AI support modes—(1) Textual Description, (2) Pros/Cons Comparison, (3) AI Scoring, and (4) Direct Recommendation—to select a final design from a set of concepts. Efficacy was measured using behavioral data (selection time, choice), user satisfaction (QUIS), cognitive load (NASA-TLX), and technology acceptance (TAM). Results show that the AI Scoring mode performed significantly better, leading to the highest user satisfaction and acceptance while imposing the lowest cognitive load. Qualitative data further suggests that while designers value the efficiency of scoring, they also demand transparency in the AI's decision-making logic.
Keywords
human-AI collaboration; design convergence; design decision-making; creative efficacy; experimental study
DOI
https://doi.org/10.21606/drs.2026.1125
Citation
Pei, Y., and Fan*, L. (2026) Evaluating creative efficacy in human-AI design convergence: An experimental study of support modes, in Simeone, L., Gray, C. M., Verhoeven, A., de Götzen, A., Bakırlıoğlu, Y., Zohar, H., Stead, M., and Buwert, P. (eds.), DRS2026: Edinburgh, 8–12 June, Edinburgh, United Kingdom. https://doi.org/10.21606/drs.2026.1125
Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License
Included in
Evaluating creative efficacy in human-AI design convergence: An experimental study of support modes
While Artificial Intelligence (AI) has demonstrated significant potential in the divergent phases of design, its role in the convergent phase (solution selection) is less understood. This study investigates how different AI-assisted selection modes impact a designer's creative efficacy. We conducted an experiment (N=40) with graduate design students, who used four distinct AI support modes—(1) Textual Description, (2) Pros/Cons Comparison, (3) AI Scoring, and (4) Direct Recommendation—to select a final design from a set of concepts. Efficacy was measured using behavioral data (selection time, choice), user satisfaction (QUIS), cognitive load (NASA-TLX), and technology acceptance (TAM). Results show that the AI Scoring mode performed significantly better, leading to the highest user satisfaction and acceptance while imposing the lowest cognitive load. Qualitative data further suggests that while designers value the efficiency of scoring, they also demand transparency in the AI's decision-making logic.